Glossary.
Key terms and concepts behind the context layer, AI context management, and how Coconut works.
AI copilot
An AI assistant that works alongside a person, helping with tasks like drafting, research, and analysis. Copilots are more effective when they draw from shared organizational context rather than relying solely on what the user provides in each prompt.
AI agent
An AI system that can take actions autonomously, such as preparing a report, triaging incoming requests, or updating a system. Agents depend on accurate, current context to make reliable decisions without constant human oversight.
Context layer
A shared foundation of organizational knowledge that every AI tool in your stack draws from. Rather than each tool starting from scratch, the context layer gives them a consistent, up-to-date picture of how your company works.
Context source
Any system, document, or data store that contributes knowledge to the context layer. Examples include CRMs, project management tools, shared drives, wikis, and communication platforms.
Context sync
The process of keeping the context layer up to date as source systems change. Context sync ensures that AI tools always draw from current information rather than stale snapshots.
Context distribution
How the context layer delivers the right knowledge to the right AI tool at the right time. Distribution is permission-aware, meaning each tool only receives the context it is authorized to access.
Context fragmentation
When organizational knowledge is scattered across multiple tools, documents, and people with no shared foundation. Fragmentation leads to inconsistent AI outputs, duplicated effort, and decisions made on incomplete information.
Context drift
When the information AI tools rely on becomes outdated or out of sync with reality. Context drift happens when there is no mechanism to keep source material current, leading to AI outputs based on stale data.
Connector
An integration that links an external tool or data source to the context layer. Connectors pull information from systems like Slack, Notion, Salesforce, or Google Drive and keep it synchronized.
Context-aware AI
AI tools that have access to relevant organizational knowledge when generating outputs. Context-aware AI produces results that are specific to your company, not just generically accurate.
Grounding
The practice of connecting AI outputs to verified, source-backed information. Grounding reduces hallucination and ensures that AI-generated content reflects your organization's actual knowledge, not general training data.
Hallucination
When an AI model generates information that sounds plausible but is factually incorrect or fabricated. Hallucination risk increases when AI tools lack access to relevant organizational context.
Institutional memory
The accumulated knowledge, decisions, and rationale that exist within an organization, often held informally by long-tenured employees. A context layer helps capture and preserve institutional memory so it remains accessible as teams change.
Knowledge base
A structured collection of information, typically articles, FAQs, or documentation. Unlike a context layer, a knowledge base is usually static and requires manual maintenance. A context layer can draw from knowledge bases as one of many sources.
Operating context
The living body of knowledge that defines how your organization operates: strategies, processes, decisions, relationships, and institutional memory. Operating context is what makes AI output relevant to your specific company, not just generically correct.
Permission-aware context
Context distribution that respects your organization's existing access controls. When context is permission-aware, AI tools only receive information that the requesting user or system is authorized to see.
Pilot
A focused, time-boxed rollout of Coconut with a single team or use case. Pilots are designed to demonstrate value quickly and establish a foundation for broader adoption.
RAG (Retrieval-Augmented Generation)
A technique where an AI model retrieves relevant documents or data before generating a response. RAG improves accuracy by grounding outputs in real information. A context layer provides the retrieval foundation that makes RAG effective at an organizational level.